IBL News | New York
According to an analysis from Sequoia Capital, only a key set of scaled players and alliances with access to vast sums of capital—including Microsoft/OpenAI, AWS/Anthropic, Meta, and Google/DeepMind—remain in play in the market for Generative AI. This report follows Sequoia’s essay “Generative AI: A Creative New World.”
The most critical model update of 2024 was OpenAI’s o1, formerly known as Q* and also known as Strawberry. This is the first example of a model with true general reasoning capabilities, which needs to stop and think before giving a response the way a human would.
• The models have largely failed to make it into the application layer as breakout products, except for ChatGPT.
• In corporate IT and global systems integrators, there are opportunities in the application layer.
Today, “wrappers” are one of the only sound methods of building enduring value. What began as “wrappers” have evolved into sophisticated cognitive architectures.
They typically include multiple foundation models with some routing mechanism on top, vector and/or graph databases for RAG, guardrails to ensure compliance and application logic that mimics how a human might think about reasoning through a workflow.
GitHub Copilot is an excellent example of this.
- Harvey: AI lawyer
- Glean: AI work assistant
- Factory: AI software engineer
- Abridge: AI medical scribe
- Sierra: AI customer support agent
- XBOW: AI-driven penetration test simulates cyberattacks to evaluate security systems
- Day.ai: AI native CRM to replace systems integrators who make billions of dollars configuring Salesforce to meet customers’ needs.
Efficient engineering is required to transform a model’s raw capabilities into a compelling, reliable, end-to-end business solution.
